损失数据自回归模型的随机梯度辨识算法

Stochastic Gradient Algorithms for AR Models with Missing Data

运用多项式变换技术将损失观测数据自回归模型转化为一种特殊形式的模型,该模型可以利用稀少观测数据来辨识模型参数,再利用模型等价原理来估计损失观测数据,提出了基于残差的随机梯度辨识算法,仿真结果验证了提出算法的有效性。

By using the polynomial transform technique, the auto-regression model with missing observation data is transformed into a special model which can be identified from scarece observation data, and the missing data are estimated by the model equivalence principle and a residual based stochastic gradient algorithm. A simulation example is included.